Grand Paris Sud Open Data (data.grandparissud.fr) — OpenDataSoft MCP. Open data for Grand Paris Sud, France via the OpenDataSoft Explore v2.1 API: urban services, mobility & environment. Keyless.
Three tools with complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have action verbs (search_, dataset_, query) and non-empty descriptions (avg 150 chars). However, parameter descriptions lack format constraints, range limits, and dependency documentation. Output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Error handling is minimal. No tool annotations (readOnlyHint, idempotentHint). Composition is sound: search→info→query workflow is logical and tools are single-purpose.
Get metadata for one Grand Paris Sud Open Data dataset (fields/schema, themes, record count) — call before query to learn the column names.
Query records from a Grand Paris Sud Open Data dataset with ODSQL. Filter (where), aggregate (group_by/select), sort (order_by), paginate (limit/offset).
Search Grand Paris Sud Open Data for datasets by keyword (urban services, mobility & environment). Returns dataset_ids (pass to query/dataset_info), titles, themes and record counts.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination) for downstream tool chaining or data extraction. search_datasets, dataset_info, and query all lack return type documentation.
Parameter descriptions lack format constraints and ranges. 'limit' and 'offset' state '1-100' and 'default 20' in text but no JSON Schema minValue/maxValue. 'where' and 'select' accept free-form ODSQL strings with no validation guidance or examples of valid syntax.
No tool annotations. Tools are read-only (safe to retry) but lack readOnlyHint in schema. Agents cannot determine idempotency or destructiveness without explicit hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Error handling is minimal. callTool() throws generic Error objects with no recovery guidance. If dataset_id is invalid or API returns 404, LLM receives raw HTTP status + truncated response text (160 chars max) with no actionable next steps.
Parameter 'query' in search_datasets and query tool is ambiguous. In search_datasets it searches titles/descriptions; in query it searches 'all fields'. Descriptions do not clarify scope or precedence vs 'where' clause.